Pan-Cancer Efficacy of Vemurafenib in BRAF^V600-Mutant Non-Melanoma Cancers

The concept that genomic alterations could be used to guide cancer therapy, regardless of tissue of origin, has been a central objective of genome-driven oncology since its inception. This initial promise was ultimately partially realized with the recent demonstration of tumor-agnostic efficacy and subsequent regulatory approval of PD-1 blockade for tumors harboring mismatch-repair deficiency (1, 2) and TRK inhibition for tumors harboring TRK fusions (3). Despite these significant advancements, experience with the broader array of targeted therapies has demonstrated that their efficacy is often dependent, at least in part, on tumor lineage (4, 5).

The efficacy of BRAF^V600 mutant–selective RAF inhibitors, including vemurafenib, provided an early and prominent example of lineage-dependent response to targeted therapy. Although these agents achieve a high objective response rate and prolong survival in patients with BRAF^V600-mutant melanoma (6), they lack meaningful single-agent activity in BRAF^V600-mutant colorectal cancer (7). Subsequent biological studies have demonstrated that the intrinsic resistance to single-agent RAF inhibition in BRAF^V600-mutant colorectal cancer is mediated by feedback reactivation of receptor tyrosine kinases (8) and may be overcome with combination targeted therapy targeting these lineage-specific primary resistance mechanisms (9). Despite these findings, the extent to which RAF inhibition efficacy in BRAF^V600-mutant tumors is conditioned by lineage beyond colorectal cancer remains largely unknown. Further complicating exploration of this important clinical question is the pattern of BRAF^V600 mutations across cancers. Specifically, although BRAF^V600 mutations are common, occurring in approximately 50% of melanomas and papillary thyroid cancers, they occur with much lower frequency (typically <5%) across a variety of other cancer types (10).

To address this important and ongoing knowledge gap, a single-arm, multi-histology, phase II study was launched (VE-BASKET; NCT01524978). This study, the first in what became a new wave of “basket” studies, was designed to assess the efficacy of a targeted agent across multiple cancer types characterized by the presence of a single genomic biomarker. Specifically, VE-BASKET was intended to explore the efficacy of vemurafenib in patients with any BRAF^V600 mutation–positive cancers other than melanoma, papillary thyroid cancer, and hairy cell leukemia, cancers for which efficacy had been previously defined in traditional tumor-specific studies (6, 11, 12). The basket study design permitted expansion or discontinuation of enrollment of any specific tumor type on the basis of observed signals of activity following initial enrollment. Previously published preliminary results from the first 95 patients who received vemurafenib monotherapy indicated promising activity in patients with non–small cell lung cancer (NSCLC) as well as histiocytic neoplasms (Erdheim–Chester disease and Langerhans cell histiocytosis; ref. 13). Expanded enrollment of patients with Erdheim–Chester disease ultimately resulted in regulatory approval of vemurafenib for this indication in the United States (14). Subsequently, more mature data in NSCLC (15), primary brain tumors (16), and myeloma (17) have been published.

Despite these tumor-specific descriptions of efficacy, the expanded experience in less commonly represented tumor types, as well as the pan-cancer activity of single-agent vemurafenib, has never been presented or reported. Importantly, during the conduct of this study, only enrollment of patients with colorectal cancer was permanently discontinued. Otherwise, accrual of patients with all other eligible cancer types was permitted throughout the study period, providing a unique opportunity to evaluate the activity of RAF inhibition in a prevalent population of patients with BRAF^V600-mutant cancers. We now present the updated and final pan-cancer efficacy data for vemurafenib monotherapy in 172 patients with 26 unique BRAF^V600-mutant cancer types. To further increase the value of our findings to the clinical and research community, patient-level demographic and efficacy data are included.

In total, 208 patients with BRAF^V600-mutant tumors were enrolled and treated. Of these, 172 had solid tumors and received vemurafenib monotherapy (Table 1; Supplementary Table S1). The remaining 36 patients excluded from this analysis comprised 9 with multiple myeloma and 27 with colorectal cancer who received vemurafenib plus cetuximab.

The median age of patients was 60 (range, 18–90) years. Patients had received a median of 2 (range, 0–10) prior lines of therapy (Table 1). The most common cancer types were NSCLC (37%), histiocytic neoplasms (16%), glioma (14%), anaplastic thyroid cancer (7%), colorectal cancer (6%), and cholangiocarcinoma (5%). A total of 26 unique cancer types were treated.

This analysis was performed after a median follow-up duration across all patients of 10.7 months (range, 0.1–46.3 months). As the study was formally closed at the time of this analysis, all patients had discontinued the study. The most common reasons for vemurafenib discontinuation were progressive disease (n = 105; 61%), adverse events (n = 20; 12%), and withdrawal by the patient (n = 13; 8%; Supplementary Table S2). In addition, 21 patients with an ongoing response or otherwise deriving benefit at the time of study closure were offered the opportunity to continue vemurafenib treatment through an expanded-access study (NCT01739764).

Vemurafenib activity is summarized in Table 2 and detailed in Fig. 1A and B. Data for patients with histiocytosis are shown in Supplementary Fig. S1A and S1B; data for patients with other non-histiocytic tumors are shown in Supplementary Fig. S2A and S2B. By investigator assessment, the objective response rate was 32.6% [95% confidence interval (CI), 25.6%–40.1%]. The best overall response (BOR) included complete responses in 5 patients (3%), partial responses in 51 patients (29.7%), stable disease of any duration in 65 patients (37.8%), and progressive disease in 35 patients (20.3%). An additional 16 patients (9.3%) had missing or nonevaluable response assessments and were counted as nonresponders per protocol. In total, responses were observed across 13 unique cancer types including NSCLC, histiocytic neoplasms, glioma of various histologies, anaplastic thyroid cancer, cholangiocarcinoma, ovarian cancer, sarcoma, salivary duct cancer, and neuroendocrine carcinoma. The clinical benefit rate (CBR), defined as a confirmed partial response of any duration or stable disease lasting ≥6 months, was 42% (95% CI, 34%–50%).

Among the 56 responding patients, the median duration of response was 13.1 months (95% CI, 8.0–22.1; Fig. 2A). Across the overall population, the median progression-free survival (PFS) was 5.8 months (95% CI, 5.4–7.6; Fig. 2B). The estimated PFS rate at 1 year was 28%. With 83 deaths (48%) at the time of study completion, the median overall survival (OS) was 17.6 months (95% CI, 13.0–28.2; Fig. 2C). The estimated OS rates at 1 and 3 years were 60% and 34%, respectively.

Adverse events occurring in ≥20% of patients, regardless of causality, are shown in Supplementary Table S3. The most common all-grade adverse events were arthralgia (45%), fatigue (34%), and hyperkeratosis (33%). In total, grade ≥3 adverse events occurred in 126 patients (73%), the most common of which were squamous cell carcinoma of the skin (15%), keratoacanthoma (10%), and maculopapular rash (9%). In addition, basal cell carcinoma of the skin occurred in 7 patients (4%) and Bowen disease in 4 (2%). Treatment discontinuation due to drug-related adverse events occurred in 13 patients (7.6%). Fatal adverse events occurred in 5 patients (pulmonary embolism, n = 2; respiratory failure and sepsis, n = 1; subdural hematoma, n = 1; and respiratory failure, n = 1). None of these were deemed related to vemurafenib.

Here we report the integrated pan-cancer efficacy of BRAF inhibition with single-agent vemurafenib in a large and diverse cohort of 26 unique cancer types harboring BRAF^V600 mutations. We found that approximately one third of patients achieved an objective response and that these responses were generally durable, with a median duration of response exceeding 1 year.

These data are noteworthy for several reasons. Our cohort included many patients with cancer types associated with poor prognosis and treatment refractoriness, including gliomas, pancreatic cancer, colon cancer, sarcomas, and cholangiocarcinomas. Despite this, confirmed objective responses were observed in 13 unique cancer types, including several in which RAF-targeted therapy remains investigational. These efficacy results were achieved despite the fact that patients with melanoma and papillary thyroid cancer were excluded from the outset, as were those with colon cancer following an interim analysis in 11 patients that indicated insufficient activity. Interestingly, a separate study of vemurafenib in BRAF^V600-mutant papillary thyroid cancer reported a response rate of 38%, similar to the response rate achieved pan-cancer here, suggesting that the exclusion of this tumor type did not bias our results (11). Beyond these limited enrollment restrictions, the cohort of patients accrued here appears to be broadly reflective of the BRAF^V600-mutant prevalent pan-cancer population. As such, these data further contribute to our understanding of the therapeutic relevance of BRAF^V600 across multiple cancer types.

When initially launching this early basket study, our primary goal was to screen for potential efficacy of RAF inhibition beyond melanoma, where this approach had already been shown to improve survival. Demonstrating the value of this approach, data from this study were used to support approval of vemurafenib in the histiocytic neoplasm Erdheim–Chester disease in the United States (14, 18), as well as inclusion of vemurafenib in the National Comprehensive Cancer Network guidelines for the management of patients with NSCLC. In parallel, alternative studies evaluating combined RAF and MEK inhibition have led to regulatory approval of such combinations in NSCLC (19) and anaplastic thyroid cancer (https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-dabrafenib-plus-trametinib-anaplastic-thyroid-cancer-braf-v600e-mutation; ref. 20). Moreover, since the conceptualization of this study, subsequent iterations of basket studies have been used to evaluate a broader range of hypotheses, including initial clinical validation of investigational genomic targets and even tumor-agnostic regulatory approval of novel molecular entities (3, 21).

This study has some important limitations. Most notably, it was launched before multiple studies demonstrated that the combination of BRAF and MEK inhibition is frequently superior to BRAF inhibition alone (22, 23). As such, the efficacy reported here may actually represent an estimate of the lower limit of what might have been achieved if this same population had been treated with a RAF–MEK inhibitor combination, although this remains unproven. Similarly, in patients with colorectal cancer, subsequent clinical data have shown that sensitivity can be induced through use of RAF inhibitor combinations that incorporate anti-EGFR mAbs (9, 24). Moreover, best response is reported as the primary efficacy endpoint, whereas in studies with a primary endpoint of response, RECIST advises that confirmed response be reported. There may therefore be a modest overreporting of efficacy as a result of this approach. The inclusion of histiocytosis in the pan-cancer analysis may also increase the reported efficacy. In addition, this study was conducted before tumor next-generation sequencing became widely available and prior to use of blood-based circulating tumor DNA sequencing. This, as well as the lack of central collection of archival tumor or plasma in the majority of patients, significantly limits our ability to interrogate the broader genomic landscape of patients treated here and our understanding of how this may contribute to the likelihood of treatment responsiveness.

In conclusion, the results of this analysis demonstrate that although tumor lineage can sometimes play an important role in conditioning response to single-agent RAF inhibition, few cancer types exhibit complete insensitivity to vemurafenib, and overall pan-cancer efficacy appears clinically meaningful. In patients with BRAF^V600-mutant tumor types for which RAF-targeted therapy, alone or in combination, is not currently approved, these data can be used to guide further discussion and decision making.

VE-BASKET was a multicenter, single-arm, phase II study of vemurafenib in patients with a variety of non-melanoma cancers harboring BRAF^V600 mutations identified through testing as routinely practiced by each participating site. All patients received single-agent vemurafenib (960 mg orally twice daily). This study was conducted in accordance with provisions of the Declaration of Helsinki and Good Clinical Practice guidelines. The protocol was approved by institutional review boards or human research ethics committees at the participating centers. All patients provided written informed consent.

Eligible patients were age ≥16 years, with histologically confirmed, measurable (RECIST version 1.1) disease, BRAF^V600 mutation–positive cancer that was refractory to standard therapy or for which standard or curative therapy did not exist or was not considered appropriate by the investigator. Patients had adequate hematologic, renal, and liver function. Prior treatment with a BRAF or MEK inhibitor was not allowed. Patients with active or untreated central nervous system metastases were excluded, as were those with melanoma, papillary thyroid cancer, and leukemia. Detailed inclusion and exclusion criteria are available in the Protocol Appendix in the supplementary material.

Assessments were performed using CT or MRI of the chest, abdomen, and pelvis at study entry and every 8 weeks thereafter until disease progression, death, or study withdrawal. Response was investigator-assessed using RECIST version 1.1. Adverse events were graded by the investigators using NCI Common Terminology Criteria version 4 (https://www.eortc.be/services/doc/ctc/CTCAE_4.03_2010-06-14_QuickReference_5×7.pdf) from consent until 28 days after discontinuation of study treatment.

The primary objective of the analysis was to evaluate the efficacy of vemurafenib (BOR). Key secondary objectives included duration of response (DOR), PFS, CBR (defined as response or stable disease of ≥6 months), OS, and safety.

This study was primarily intended to analyze efficacy in a tumor-specific context. As such, the original statistical design utilized a modified, two-stage Simon design study for each prespecified tumor cohort (NSCLC, ovarian cancer, colorectal cancer, cholangiocarcinoma, breast cancer, and multiple myeloma). During stage 1,7 patients with measurable disease were enrolled into each cohort; this stage was considered complete when all patients had completed a minimum of 8 weeks' treatment, developed progressive disease, prematurely withdrew, or died. A further 6 or 12 patients could be enrolled into stage 2, depending on the results for stage 1: If 2 to 4 of the 7 patients had a response, an additional 12 patients could be enrolled; if 5 or more of the 7 responded, 6 additional patients were recruited. Recruitment into any cohort/indication could be further expanded up to 70 patients if a response rate was demonstrated in stage 2 for that cohort according to protocol-defined stopping rules or a clear clinical benefit was observed, as determined by the steering committee.

In this analysis, efficacy data were pooled across all solid tumor patients treated with single-agent vemurafenib with the goal of defining pan-cancer efficacy. DOR, PFS, and OS were calculated using the Kaplan–Meier method. CIs for proportions were calculated using the Clopper–Pearson method. All analyses were performed using SAS (versions 9.2 and 9.4; SAS Institute Inc.).

V. Subbiah is a scientific advisory board member at Helsinn, Loxo Oncology/Eli Lilly, R-Pharma US, INCYTE, QED, Medimmune, and Novartis, reports receiving commercial research grants from Roche/Genentech, Novartis, Pharmamar, Inhibrx, Exelixis, Blueprint Medicines, D3, Pfizer, MultiVir, Amgen, AbbVie, Alfa-Sigma, Bayer, Agensys, Boston Biomedical, Boston Pharmaceuticals, Turning Point Therapeutics, Idera, Loxo Oncology/Eli Lilly, Medimmune, Altum, Dragonfly Therapeutics, Takeda, GSK, Nanocarrier, Vegenics, Northwest Biotherapeutics, Berg Health Pharma, Incyte, and Fujifilm, and has received other remuneration from Medscape, ASCO, and ESMO. I. Puzanov is a consultant at Amgen and has ownership interest (including patents) in Celldex. J.-Y. Blay reports receiving commercial research support from Roche Genentech, Novartis, and GSK, and has consultant/advisory board relationships with Novartis and Roche Genentech. I. Chau is an advisory board member at Bristol-Myers Squibb, Eli Lilly, MSD, Bayer, Roche, Merck Serono, Five Prime Therapeutics, AstraZeneca, Oncologie International, and Pierre Fabre. J. Wolf is an advisory board member for and has received a lecture fee from AbbVie, Amgen, MSD, Novartis, Pfizer, Roche, Takeda, AstraZeneca, BMS, Chugai, Janssen, Lilly, and Loxo, is an advisory board member at Blueprint, has received a lecture fee from Boehringer Ingelheim, and reports receiving commercial research grants from BMS, Johnson and Johnson, and Novartis. J. Baselga is EVP, Oncology R&D, at AstraZeneca, is a member of the board of directors at Foghorn, Varian, Bristol-Myers Squibb, Grail, Aura, and Infinity, is a consultant at PMV Pharma, Juno, and Seragon, reports receiving a commercial research grant from Roche, and has ownership interest (including patents) in Apogen, Tango, and Venthera. F. Meric-Bernstam is chair, Department of Investigational Cancer Therapeutics at MD Anderson Cancer Center, is a consultant at eFFECTOR Therapeutics, Samsung Bioepis, Aduro BioTech, Kolon Life Science, OrigiMed, Sumitomo Dainippon Pharma Co., Seattle Genetics Inc., Debiopharm, Dialectica, Piers Pharmaceuticals, Xencor, PACT Pharma, Zymeworks, Jackson Laboratory, Genentech, F. Hoffman-La Roche, Parexel International, Pfizer, and IBM Watson, reports receiving commercial research grants from Aileron Therapeutics, AstraZeneca, Guardant Health, K Group, Millennium Pharmaceuticals, Novartis, PPD Investigator Services, Puma Biotechnology, Seattle Genetics, Taiho Pharmaceutical, Zymeworks, Bayer Healthcare Pharmaceutical, Calithera Biosciences, Curis Inc., CytomX Therapeutics, Daiichi Sankyo, Debiopharm International, eFFECTOR Therapeutics, and Genentech Inc., has received speakers bureau honoraria from Sumitomo Dainippon Pharma and Dialectica, and has advisory board relationships with Puma Biotechnology, Mersana Therapeutics, Immunomedics, Silverback Therapeutics, Inflection Biosciences, Spectrum Pharmaceuticals, Seattle Genetics, Grail, Darwin Health, and Clearlight Diagnostics. E.L. Diamond has an advisory board relationship with Third Rock Ventures. G.J. Riely reports receiving commercial research grants from Roche, Takeda, Mirati, Pfizer, and Merck, and has advisory board relationships with Takeda, Merck, and Daiichi. E.J. Sherman is an advisory board member at Loxo, Regeneron, Novartis, and Eisai. T. Riehl is a clinical scientist at Roche/Genentech. B. Pitcher is a senior statistical scientist at Hoffmann-La Roche Limited. D.M Hyman is chief medical officer at Eli Lilly/Loxo Oncology, is a consultant at Chugai Pharma, Boehringer Ingelheim, Pfizer, Bayer, Genentech/Roche, and Eli Lilly, is a scientific advisory board member at Kinnate, and reports receiving commercial research grants from AstraZeneca, Puma Biotechnology, Loxo Oncology, and Bayer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Subbiah, I. Puzanov, J.-Y. Blay, I. Chau, N.S. Raje, J. Baselga, G.J. Riely, D.M. Hyman

Development of methodology: V. Subbiah, I. Puzanov, J. Baselga, T. Riehl, D.M. Hyman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Subbiah, I. Puzanov, J.-Y. Blay, I. Chau, A.C. Lockhart, J. Wolf, J. Baselga, F. Meric-Bernstam, E.L. Diamond, G.J. Riely, E.J. Sherman, D.M. Hyman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Subbiah, J.-Y. Blay, I. Chau, N.S. Raje, J. Wolf, J. Baselga, F. Meric-Bernstam, J. Roszik, G.J. Riely, E.J. Sherman, B. Pitcher, D.M. Hyman

Writing, review, and/or revision of the manuscript: V. Subbiah, I. Puzanov, J.-Y. Blay, I. Chau, A.C. Lockhart, N.S. Raje, J. Wolf, J. Baselga, F. Meric-Bernstam, J. Roszik, E.L. Diamond, G.J. Riely, E.J. Sherman, T. Riehl, B. Pitcher, D.M. Hyman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Subbiah, I. Puzanov, D.M. Hyman

Study supervision: V. Subbiah, I. Puzanov, J.-Y. Blay, A.C. Lockhart, J. Baselga, G.J. Riely, T. Riehl, D.M. Hyman

This study was supported by F. Hoffmann-La Roche Ltd. Additional support was provided by the NIH (P30 CA008748). J.-Y. Blay was funded in part by grants [NETSARC+ and LYRICAN (INCA-DGOS-INSERM 12563)]. Editorial support was provided by Lee Miller and Deirdre Carman, PhD, of Miller Medical Communications Ltd. funded by F. Hoffmann-La Roche Ltd.